Objective: The aim of this study was to establish a predictive model for Down's syndrome using maternal age as well as maternal serum levels of alpha-fetoprotein (AFP) and human chorionic gonadotropin (hCG), and to identify an optimal cut-off risk in women under the age of 35 years to improve sensitivity.
Methods: Logistic regression models were utilized to predict fetal Down's syndrome as a function of maternal age and logarithm of levels of AFP as well as hCG using training data of 20 pregnancies with fetal Down's syndrome and 9730 unaffected pregnancies. Validation was performed using data of another nine affected pregnancies and 3496 unaffected pregnancies. Receiver operating characteristic (ROC) curves were plotted.
Results: Based on the newly established logistic regression equations, the optimal cut-off risk from the ROC curve analysis was at 1:499, with a 17.8% false-positive rate and a 90.0% sensitivity. A suboptimal cut-off risk was estimated at 1:332, with a 12.0% false-positive rate and an 80% sensitivity.
Conclusion: A predictive model for Down's syndrome was developed using logistic regression. By ROC curve analysis and clinical consideration, the cut-off risk for young pregnant women could be determined.